Design and operate data centres for the AI era

Most data centres currently in operation were not designed to support the high-power requirements of artificial intelligence (AI)-led computing workloads. New infrastructure requirements are different from traditional data centres, as they generate greater levels of heat that current facilities cannot remove fast enough.

According to an article published in DataCentre Magazine, for typical enterprise applications, a data centre might require 7-10 kilowatts (kW) of power per rack ^[1]. But for AI, the power requirement increases to more than 30kW per rack. Consequently, data centre campuses must be upgraded to accommodate the digital infrastructure associated with AI workloads as well as to allow for significant cooling systems and power distribution units (PDUs), generators, and uninterruptible power supplies (UPS).

 

According to IIR latest technical brief, “Cooling technologies for sustainable data centres”, artificial intelligence also opens a new field of R&D to manage the daily operation, predict cooling loads and optimise performance. Indeed, data centres operate continuously throughout the year, and the operating loads of the IT equipment and outdoor weather conditions are constantly changing, meaning that efficient operation is important to achieve energy conservation and emission reductions.

There is still much work to be done before AI can optimally control data centres operation. Existing machine learning control methods represent only preliminary attempts to optimise energy efficiency and are still at the pilot/test stage.

 

When building data centres that can support higher computing workloads, the industry focuses on regions and climates that can provide benefits such as free cooling or renewables to decarbonise. In Nordic countries for instance, the climate allows for more energy efficient cooling technologies such as natural air cooling or Direct Liquid Cooling (DLC).

These techniques significantly lower the Power Usage Effectiveness (PUE) of data centres, a 33% improvement in energy efficiency and significant carbon reductions, particularly when combined with use of renewable energy, such as that of Iceland ^[1].

 

$$PUE\; =\frac{Total\; Facility\; Energy\; Usage}{Energy\; Usage\; by\; IT\; equipment}$$

 

 

• A PUE of 1.0 indicates perfect energy efficiency, where all the energy used is consumed by the IT equipment, and none is wasted on supporting infrastructure such as cooling, lighting, or power distribution.
•  Typical PUE values for modern data centers are generally between 1.1 and 1.8, where values closer to 1 are considered better, indicating that less energy is wasted in cooling

 

In simple terms, the lower the PUE, the more energy-efficient the data center is. Reducing PUE is a primary goal for data center operators aiming to improve sustainability and reduce operational costs.

 

In addition, liquid-cooled data centres have a higher potential for waste heat recovery due to their higher temperatures of waste heat, according to IIR ^[2]. Recycling waste heat from internet data centres is a promising way to reduce costs and increase efficiency.

 

 

 

 

Sources

[1] Jackson, A. (2024, November 29). How to Design and Build a Data Centre for the New AI Era. https://datacentremagazine.com/articles/how-to-design-and-build-a-data-centre-for-the-new-ai-era

[2] Shao S., Zhang P., Li X. Cooling technologies for sustainable data centres. 59th Technical Brief on Refrigeration Technologies. International Institute of Refrigeration (IIR), Paris. http://dx.doi.org/10.18462/iir.TechBrief.01.2025